Behave, or else?
Aid and voting on the UN Security Council
Svanhildur Thorvaldsdottir∗
†
August 15, 2015
Work in Progress. Please do not cite without permission.
Abstract. Recent scholarship has shown that states holding a temporary seat on
the United Nations Security Council receive higher levels of aid from the United
States. This has led to the question: are states being bribed for their votes?
The results from the existing literature support an affirmative answer to this
question. Using recently available data on votes in the UN Security Council, I
address this question more directly than previous studies have been able to. I
find that the US does use financial incentives to induce states to vote with them
on the Security Council. However, on average, the US does not seem to punish
states (or reward them) for their voting behavior. This suggests that the US may
not be bribing for votes so much as buying goodwill from the elected members of
the Security Council. These findings add greater nuance to the current analysis
of vote-buying within international organizations, and can also serve as a starting
point for further research into the behavior of the members of the UN Security
Council.
∗
For comments and suggestions on earlier versions of this paper, many thanks to Randy Stone, Rabia
Malik, Chitralekha Basu, Kevin Clarke, Hein Goemans, Stuart Jordan, Brenton Kenkel, Matthew Knowles,
William Spaniel, and Susanna Supalla. All errors remain my own.
†
University of Rochester. [email protected]
1
Introduction
When Iraq invaded Kuwait in 1990, one of Iraq’s long-standing allies, Yemen, happened to
hold a seat on the UN Security Council. In a story that has subsequently become part of
Security Council lore, Yemen refused to cast an affirmative vote on what became Security
Council Resolution 678, which authorized UN member states to “use all necessary means to
uphold and implement resolution 660 (1990) and all subsequent relevant resolutions and to
restore international peace and security in the area” (United Nations, 1990). The United
States was deeply dissatisfied with this decision by Yemen, and, in response, decreased its
aid to the country from a little over USD 68 million in 1990 to approximately USD 12 million
in 1991.
The above anecdote indicates that states—including the United States—care about the
UN Security Council. Indeed, how states behave while members of the Security Council
seems to be quite important at times, and a healthy body of literature has developed that
discusses the Security Council and its effects on world affairs. A strand of this literature
has focused on the role of the 10 temporary members of the Council—and in particular on
whether these members stand to gain from their tenure. The relationship between membership and US aid, the number of conditions attached to International Monetary Fund
(IMF) programs, and the number of World Bank programs have all been estimated, with
the general result that Council membership makes states considerably more likely to receive favorable treatment by the US and by organizations in which the US wields significant
influence (Kuziemko and Werker, 2006; Dreher, Sturm and Vreeland, 2009a,b).
A few hypotheses have been developed to explain this apparent association between
membership on the UN Security Council and increased aid from the United States or other
2
benefits. The first has already been touched upon, and posits that that the US bribes Council members using either its own aid or its influence within other international institutions to
secure benefits for well-behaved states, while also punishing states for recalcitrant behavior.
Another suggests a “limelight” effect, i.e., that states are able to draw greater attention to
their needs while they sit on the Council and that this explains why their aid increases during
their tenure. A third hypothesis claims that an unobserved variable measuring “international
involvement” is associated with both increased aid receipts and an increased probability of
Council membership, and that this intervening variable is what drives the observed relationship, rather than vote-buying. A fourth possible explanatory mechanism, on which the
literature has hitherto remained silent, is that rather than directly bribing countries, the US
buys votes by buying goodwill from other states. That is, the US, seeking to increase the
likelihood that states with vote with it on the Council, increases aid to states while they
are on the Council. In this explanation, the US is not directly bribing states for individual
votes, but is buying a “friendly” Council.
To date, however, this body of research has been unable to account for actual state
behavior while on the Council. That is, it has been unable to answer the question of whether
this posited bribery is taking place, and, if it is, whether the bribery actually works. This
paper uses a recently compiled dataset with voting records for all Council members on all
resolutions and vetoes to address this question more directly, allowing us to parse out more
clearly which of these possible mechanisms is likely at play.
Unfortunately for the analyst, aid is not committed directly following (or in advance of)
each vote. Rather, aid is reported on an annual basis by the USAID. Thus, any analysis of
voting and aid must be done at the year level. As votes in the Security Council take place
throughout the year, some clearly take place before aid is committed in that year, and some
3
after, which makes aid and votes endogenous to one another. By explicitly modeling this
endogeneity using a three-stage least squares estimator, I permit aid and votes to affect each
other simultaneously. My overall results show that aid does have a significant impact on how
states vote while on the Security Council, but that voting does not have the same impact on
aid. This finding detracts from the “limelight” or “international involvement” hypotheses,
as those would predict a relationship between Council membership and aid, but not between
votes and aid, and suggests that vote-buying is, in fact, taking place.
Which vote-buying mechanism is driving the results—bribery or goodwill purchase—is
hard to separate out empirically, as they are likely to be observationally equivalent. However, there are reasons to expect that the goodwill mechanism is at play rather than the
bribery one. First, the US prefers consensus on the Council (Dreher and Vreeland, 2011).
Unanimity lends greater credibility and legitimacy to Council actions, which sits well with
the US domestic audience (Fang, 2008; Chapman and Reiter, 2004). Second, having consensus authorization from the Security Council for military interventions may increase the
likelihood that states actually contribute sufficient numbers of peacekeepers to particular
missions. Third, the US may not always be able to credibly threaten countries if they fail
to comply with its wishes, which may render bribery impossible. Fourth, unless the US is
simply extraordinarily good at bribing, and thus virtually never has to punish countries for
noncompliance, it seems likely that there should be a discernible effect of votes on aid. Given
that this finding is not supported by the data, after agreeing with the literature in rejecting
the “limelight” and “international intervention” hypotheses, I tentatively conclude that the
US is likely buying votes by buying goodwill among others, rather than bribing countries for
their votes.
This paper proceeds by first giving an overview of how the UN Security Council operates
4
and how states are elected to it. It then reviews the literature that has addressed vote-buying
in international institutions to date, and discusses how the hypotheses relating to the Security
Council developed by this literature cannot be adequately answered in the absence of actual
voting data. In addition, it discusses some possible reasons for why states might be willing
to sell their votes, and subsequently develops hypotheses about the relationship between aid
and votes that we should observe if bribery is, in fact, taking place. The following section
analyzes the results from the statistical estimation of the relationship between aid and votes,
and concludes by discussing potential future avenues for research.
2
United Nations Security Council
2.1
Why Security Council membership?
Widely considered the central organ of the United Nations, the Security Council has garnered
the most attention from policy makers and scholars alike. In large part, this follows from
the unique privilege awarded members of the Security Council by Article 24.1 of the UN
Charter: the ability to pass resolutions that are binding on the organization’s member states.
Consequently, a position on the Security Council is widely viewed as more influential and
important than positions within other UN bodies. Furthermore, to the extent that Security
Council approval for military operations or economic sanctions is seen as enhancing the
legitimacy of such endeavors, the Council can have important influence on world affairs.
Thus, the Council’s fifteen members have considerable influence beyond that of most other
UN member states.
There is considerable literature available on the functioning of the Security Council and
5
how it has affected world affairs generally, as well as its impact on particular issues or
operations (Luck, 2006; Malone, 2006b,a; Lowe et al., 2008). Unsurprisingly, much of this
literature has focused on the role of the permanent members—the United States, the United
Kingdom, France, Russia, and China. These countries, in particular the United States, wield
the greatest influence on the Council, and much of the Council’s history can be written as
a story of superpower relations. Thus, in light of the marginal—and maybe even purely
procedural—effect of the elected members of the Council, it is somewhat unclear why states
would seek non-permanent seats.
Scholars have proposed a range of explanations for why states might seek Council membership. For instance, Hurd puts significant stock in the symbolic authority of the UN
Security Council and argues, “Council membership confers status and recognition on a state
and allows the state to appropriate some of the authority derived from the legitimacy of the
Council” (2002, p. 43). He furthermore notes that the cost of elections has risen quite a
bit in recent years (Hurd, 2002, p. 43), which provides evidence that states do believe that
membership is valuable. Malone (2000) largely concurs, but adds that states may run either
to advance a particular national cause (such as Morocco on Western Sahara in in the early
1990s) or to further a broader agenda or issue that the state cares about (such as Canada’s
stated agenda to promote human security during its 1998 Council bid).
Another strand of literature examines the hypothesis that states stand to gain financially
from their membership on the Council. Kuziemko and Werker (2006) show that Council
membership is associated with a 59 percent increase in total aid from the United States.
They further show that, in diplomatically important years, this estimated effect rises to
almost 170 percent.1 They also find a statistically significant, but smaller, effect of Secu1
Diplomatically important is defined as years in which mentions of the United Nations Security Council
6
rity Council membership on UN aid, and show that this appears to be mostly driven by
UNICEF. Dreher, Sturm and Vreeland (2009a) find that Security Council membership has
a significantly positive effect on the number of new World Bank projects approved. Dreher,
Sturm and Vreeland (2009b) also find a positive and significant relationship between UN
Security Council membership and participation in IMF programs and a negative and significant relationship between the number of conditions attached to said programs. These results
are notable, not least because all the international organizations mentioned— UNICEF, the
World Bank, and the IMF—are known to be heavily influenced by the United States.
The literature examining the relationships between Security Council membership and
various forms of aid or preferential treatment has posited three main hypotheses for its
causes. First is the hypothesis that Council membership enables countries to better bring
attention to their needs on the world stage; i.e., they are in the international “limelight,”
so to speak (Kuziemko and Werker, 2006; Dreher, Sturm and Vreeland, 2009a,b). Kuziemko
and Werker (2006) note that aid receipts increase sharply while countries are members of
the Council, and return to their previous levels shortly thereafter. This, they maintain,
undermines the hypothesis that states are simply better able to draw attention to themselves
while on the Council, because if that were the case, aid receipts should stay high even after
countries’ Council tenure.
However, it is not clear why being in the international limelight for a couple of years
should lead to a permanent increase in aid receipts from the US. The US gives aid to a huge
number of countries, and presumably shifts its aid focus depending on where its priorities
lie from year to year. Thus, an average decrease in aid receipts post-Council membership is
certainly not sufficient evidence to reject the limelight hypothesis. Indeed, this hypothesis
in the New York Times exceed a certain threshold.
7
would also be consistent with decreased aid receipts following Council membership: once
states are off the Council, their ability to attract attention diminishes, while that of other
countries (that replace them on the Council) increases. Aid may thus decrease simply as
a function of the fact that donors have limited funds to hand out, and limited time to
determine who is most in need, while at the same time spending considerable amounts of
time interacting with a relatively small set of developing states.
The second hypothesis that has been put forth is essentially that of spurious correlation;
i.e., that increases in aid receipts go hand in hand with an increased probability of serving on the UN Security Council, while both are caused by a third variable, which might
be termed “international involvement.” Scholars largely reject this hypothesis for a similar
reason as the first one. Kuziemko and Werker (2006) note that aid receipts tend to increase
dramatically during Council membership and argue that if it were being caused by greater
international involvement, this upward trend in aid should be much smoother than what is
actually observed empirically.
Again, it is unclear that increased international involvement should necessarily be associated with a smooth increase in aid receipts. For example, states often become more
involved in international organizations as they become richer, which should be associated
with decreased—not increased—aid receipts. In addition, there is no reason to assume that
states’ international involvement should increase or decrease smoothly, as largely exogenous
factors (such as conflict or drought) can cause it to change quite suddenly. Furthermore,
there is little reason to expect the probability of membership on the Security Council to
increase purely as a function of states becoming more involved internationally. Indeed, some
former members of the Council have not actually run for seats, but rather have been elected
as compromise candidates after several rounds of unsuccessful voting in the General As8
sembly (this was the case with Panama in 2006). In these cases, their membership status
says little about their overall international involvement. Further, many countries that have
very low profiles within most international organizations have been members of the Security
Council. This is particularly true of many African countries that have sat on the Council
mostly by virtue of the rotational system that the African Union has put into place. It thus
seems rather too demanding of the data to expect that both aid receipts and the probability
of Council membership increase smoothly and at roughly the same pace, which is what this
hypothesis requires.
The third hypothesis, and the one the literature has converged on as the most likely
explanatory mechanism, is that the US is buying votes from Council members. Kuziemko
and Werker (2006) and Dreher, Sturm and Vreeland (2009a,b) argue that it is not simply
the case that the United States is rewarding countries for sitting on the Council (which can
be a costly venture for states, at least financially and sometimes politically as well), but that
the US is actively buying their votes on the Council. This, the argument goes, explains why
aid increases dramatically when countries join the Council and decreases shortly after their
tenure.
As was noted above, there are two different ways in which vote-buying can take place.
First, the US could be bribing countries for their votes; that is, there is some sort of mutually
understood arrangement that the US will give a country greater amounts of aid in exchange
for that country voting more with the US during its membership term. If the deal is broken,
the US will likely retaliate by giving less aid. Second, rather than making an explicit deal
with other countries, the US could be buying votes by buying goodwill; by increasing the
amount of aid that it gives to other countries, the US makes itself better liked, and therefore
increases the likelihood that others will vote with it on the Council.
9
2.2
Why vote-buying?
Regardless of the method of vote-buying employed, it is necessary to first explore why the US
would be at all interested in buying the votes of other states on the Security Council. The
argument that vote-buying takes place on the Security Council through aid to temporary
members requires three main components. First, aid is a strategic tool, i.e., it is not purely
driven by recipient need, but also by donor preferences. This argument is not new in political
science and should require little explanation. For example, discussing donors more generally,
but with special attention given to the US, Alesina and Dollar (2000) show that strategic
considerations determine aid flows at least as much as the needs and policy performance of
recipients. This follows on earlier literature that has made a similar argument about the
strategic element of US aid giving (cf. Morgenthau, 1962; McKinlay and Little, 1977).
Second, it must be the case that, at least some of the time, the United States cares
about outcomes on the Security Council; in particular, it must care about how the elected
members vote. There is reason to believe that this may actually be the case. Sandler and
Harley (1999) note, “when acting through the UNSC, the US bears a smaller share of military
burdens and needs fewer soldiers as compared to unilateral actions” (cited in Dreher, Sturm
and Vreeland, 2009a, p. 2). Thus, even though the US may increase aid payments to some
members to help secure a better functioning Council, it may still end up with a net gain, as
compared to the potential costs of having to intervene unilaterally with greater frequency.
In addition to lower financial obligations from having to intervene on its own, the US may
also prefer to have the backing of the international community, through the sanction powers
of the Security Council. Indeed, it has been shown that, within the US, public legitimacy of
military action is higher when the Security Council’s approval is obtained and, furthermore,
10
that the calculus of all relevant actors may change as a function of Council authorization
(Fang, 2008; Chapman and Wolford, 2010).
The third necessary component of the hypothesis that the United States buys the votes of
temporary members of the Security Council is that the elected members have to be responsive
to the vote-buying efforts, at least on average. If the elected members of the Council were
unbribable, the US would presumably have noticed this at some point and stopped trying.
Thus, in theory at least, the system constitutes a self-reinforcing equilibrium of sorts, where
the US obtains what it wants when it really matters, elected members are able to enjoy a
few spoils of membership, as well as the prestige of serving on this eminent global body, and
the Council looks like it operates as the deliberative body it was made to be. This is not
to say that the system always works like a well-oiled machine. The snags are usually also
quite well-publicized by the international media, so there is little room for the illusion that
everyone is always on the same page with how the Council is ‘supposed’ to operate. The
question remains open, however, whether what scholars think is going on within the Security
Council chambers is actually what happens in practice.
Although only indirectly related to the Security Council, a parallel and relevant body
of literature has investigated the vote-buying hypothesis in the context of the UN General
Assembly, another one of the main organs of the UN and widely considered its main deliberative body. Although the greatest support for the vote-buying hypothesis has been uncovered
when examining only those votes in the General Assembly that the US government has defined as important, Dreher, Nunnenkamp and Thiele (2008) find evidence of vote-buying
in the General Assembly by the US even when examining all votes. They also look for
vote-buying attempts from other G7 members, but do not find a similar patterns. Using a
subset of only important votes in the General Assembly (as defined by the US government),
11
Carter and Stone (2011) find evidence that the US targets countries selectively when trying
to buy influence, focusing its efforts on democracies, which, they find, are more responsive
to such attempts. They furthermore find that the US punishes and rewards aid recipients
for General Assembly votes differently depending on their levels of development, political
orientation of their governments, and alliance relationships (p. 4).
The body of literature that has engaged the question of vote-buying in the context of
the Security Council has, to date, focused on the first two components of the aid-for-votes
hypothesis, namely that aid is a strategic tool of foreign policy and that the US cares
about how states vote. It has been demonstrated repeatedly that aid is not only used for
its purported ends, but also for strategic purposes. Anecdotal evidence, theories about
international governance, and empirical evidence supporting strategic aid to elected Security
Council members have been heralded in support of the second component. However, when
it comes to whether states are actually responsive to vote-buying, scholars have been silent.
The literature largely assumes that states are responsive, but there is little direct evidence
to support that belief.
Without the ability to address this third component, i.e., that states are responsive to
the vote-buying efforts of the US, the aid-for-votes hypothesis has remained inadequately
addressed. In particular, without voting data researchers have been prevented from fully
distinguishing between the three competing hypotheses explaining the increase of aid during
Security Council tenure discussed above: namely, whether the explanation is vote buying
(and then what kind of vote-buying), whether states are simply able to draw greater attention
to their development needs during their two-year stint in the international limelight, or
whether the trend is due to spurious correlation. It is the contribution of this paper to begin
to address this third component by analyzing the voting behavior of the elected members of
12
the UN Security Council and its relationship with US aid, using a recently compiled dataset
of Council roll-call votes.
2.3
Why vote-selling?
As in all goods markets, where there is demand, there is also supply. Thus, it is not enough
that the US seems to be willing to buy votes of elected Security Council members; it must, in
fact, be the case that members are willing to sell. And given the fervor with which many UN
members talk about the need for legitimacy, transparency, and “the community of nations,”
the idea that their votes might be for sale when they gain seats on the Security Council does
require some discussion. In short, why should elected members be swayed by US influence?
The first part of the answer is simply that sometimes they are not swayed. Yemen is
a case in point. However, the Yemen example, interesting as it may be, is probably not
representative of the average behavior of elected members of the Council. Secondly, and
what might be the most plausible explanation, those elected members that are low-income
countries need money that the US is willing to hand out. In particular, they care about aid
money more than they care about ideals of transparency and legitimacy at the UN.
The third part is that states know that the US has a credible outside option; it can always
act outside the UN, should it deem an issue of sufficient importance (cf. Stone, 2011). Thus,
a choice that elected members of the Security Council might plausibly be faced with is (1)
accept increased aid payments from the US while on the Council in exchange for generally
behaving the way the US wants, or (2) reject increased aid payments (or, worse still, accept
them and not fulfill their end of the bargain, whether or not that bargain was made explicit).
Those choosing the second option would suffer retaliation from the US for their recalcitrant
13
voting behavior and risk making the Council look ineffective and weak when the US decides
to ignore its views. It is likely that most countries, most of the time, would pick the former
option.
What states probably want to avoid, however, is being seen as either accepting bribes for
their votes or as condoning such behavior. This goes for the US as well—it would rather not
be caught in its vote-buying. Therefore, the explicit references to costly votes that have made
the Yemen example so prominent are likely to be quite rare. Thus, we may not observe the
US making aid decisions in response to particular votes; instead, the US may be effectively
trying to buy good behavior by buying goodwill.
2.4
Hypotheses
If the causal mechanism that the literature has converged upon, namely vote-buying, is the
true mechanism, then it should either be the case that increased aid from the US leads to
an increase in the number of times a state votes with the US while on the Security Council
or that voting more with the US while on the Security Council leads to increased aid from
the US.
If both of these hypotheses are rejected, it stands to reason that something other than
vote-buying is explaining the finding in the literature that states receive more aid while on
the Security Council. Such a finding would lend greater support to either the “limelight”
effect or the “international involvement” story. Distinguishing between these two hypotheses,
however, is not something that the current analysis is capable of, as addressing that question
would require a considerably different set of variables.
Although one might think that the US would simply seek to build a minimum winning
14
coalition in order to pass the resolutions it wants, which means that neither of the above
hypotheses necessarily holds up to empirical testing, there are a few reasons why one might
still expect to see this effect of vote-buying. First, the US may seek to buy as many votes as
it can as a form of insurance. Dreher, Sturm and Vreeland (2009b) note that the vote-buying
literature has established that states seek to build over-sized coalitions (p. 744). Second, if
the US is buying votes by buying goodwill, then it would likely increase aid to all (or nearly
all) temporary members of the Security Council, rather than just to a select few. This
mechanism also supports the prevailing view that unanimous (or near unanimous) votes are
preferred on the Council (Dreher and Vreeland, 2011) but does so for somewhat different
reasons than previous literature has claimed. Another reason why the US may seek to buy
votes from all of the temporary members is that it does not want to be seen as bribing states
for their votes, so it tends to seek agreement on everything. Similarly, states do not want
to be seen accepting bribes from the US, nor do they wish to see a significant decrease in
their aid receipts as a results of having gone against the wishes of the US, so they tend avoid
rocking the boat much during their tenure on the Council.
3
3.1
Data and methodology
Data
In order to evaluate the hypothesis that voting on the UN Security Council affects aid
receipts from the US, I construct a panel dataset of all Security Council members from 1946
to 2007—the time period for which roll-call data is currently available. This gives me a
total of 850 possible observations. I exclude from my analysis the five permanent members
15
of the Council as it is likely that their voting behavior would be qualitatively different
from that of the elected members, by virtue of their veto power. This leaves me with 540
observations of 113 distinct countries. In order to avoid muddying the results of the analysis
by including countries not in need of US aid, I also remove high-income countries from the
dataset. Following Kuziemko and Werker (2006) and the subsequent literature, I define a
high-income country as one that was classified as such by the World Bank in 2003.2 This
further reduces my number of observations to 395, with 89 distinct countries.
The recently-coded data on roll-call votes for the UN Security Council comes from Dreher
and Vreeland (2011). For every state that was a member of the Security Council in a given
year, I calculate the percentage of times that it voted with the United States in that year.
Following the convention of the literature on voting in the UN General Assembly, I code a
state’s vote with the US as 1, a vote against the US as 0, and abstentions, when the US
either voted yes or no, as 0.5 (Thacker, 1999). Yes or no votes when the US abstains are also
coded as 0.5.3 Absences (extremely rare phenomena in the data) are dropped. As Figure 1
indicates, in most years, the Security Council has around 15-25 votes. When plotted over
time, however, it becomes clear that the number of votes per year increases dramatically
after the end of the Cold War. This comports with historical evidence that Council activity
suffered considerably due to superpower rivalry during the Cold War.
It should be noted that the roll-call votes only include successfully passed resolutions
(consensus or not), as well as vetoed resolutions. It includes only partial information on the
2
A possible drawback to this restriction is that it excludes countries that are classified as high-income in
2003, but perhaps should not have been classified as high-income in earlier periods. An examination of the
GDP of countries listed as high-income did not reveal any cause for concern that relevant country-years had
been excluded.
3
As an additional check, I also coded a variable where abstentions are coded as 0 and find the same
substantive results.
16
Figure 1: Descriptive statistics, SC votes
rare failed-majority vote, which occurs when a resolution fails to obtain the required 9 votes
for passage (7 votes prior to Council expansion). As these votes are quite rare, I do not
believe that excluding them significantly impacts my analysis.
US foreign aid data was obtained from the “Greenbook,” the U.S. Overseas Loans &
Grants database compiled by USAID. Descriptive statistics for this variable, along with
the other main variables used in my empirical analysis, are presented in Table 1. The
“Greenbook” reports both economic aid and military aid, and I sum and log the two to
create my aid variable. Of all the country-years in the dataset, almost 90 percent received
economic aid and about two-thirds received military aid. The data are reported in constant
2010 US dollars.
It is the case, however, that aid commitments have to be approved by the US Congress.
Therefore, it is likely to be the case that there is some lag between the time a country is
promised aid and the time that that aid is actually officially reported by the US government.
Due to this, I lag all my variables by one year, except for the aid variables. This allows me
17
to reflect the reality that aid committed in a given year was probably decided upon by the
executive branch of the US government in the previous year.
As control variables, I employ a governance measure, an economic measure, a population
control, a military alliance portfolio measure, and country fixed effects. The governance
measure is intended to control for similar governance structures making states more likely to
vote together on the Security Council, and it also helps control for the extent to which the
US may use governance features to help determine its aid commitments. The variable I use
is the Polity score, measured using Polity IV’s 21-point scale, where -10 indicates complete
autocracy, and 10 indicates complete democracy (Monty G. Marshall and Jaggers, 2010). As
an economic control, I used logged values of real GDP per capita, obtained from the Penn
World Table (Heston, Summers and Aten, 2011). Increased GDP should be associated with
decreased need for aid, so that measure will control for the extent to which aid is not a helpful
tool for vote-buying. The Penn World Table reports GDP figures from 1950 onwards, but
the start-date varies a bit from country to country. Military alliance similarity with the US is
measured using the S score calculation of similarity between alliance portfolios (Bennett and
Stam, 2000). The literature on bribery on the Security Council does not routinely control
for military alliances; however, despite using military aid as part of the dependent variable
(in addition to economic aid). As the last variable is not a routine control in this literature,
I include regressions both with and without this measure.
3.2
Modeling endogeneity
There is reason to believe that the two main variables of interest—aid and votes—are endogenous to one another. While there is anecdotal evidence that the US has punished states
18
Table 1: Descriptive Statistics
Variable name
Minimum
Median
Mean
Maximum
Votes with US
(log)Aid from US
Polity
(log)GDP per capita
(log)Population
S score
57.14
0.00
−10.00
5.36
12.78
0.00
91.38
17.57
−4.00
7.79
16.36
0.37
87.19
16.57
−1.05
7.78
16.36
0.43
100.00
22.25
10.00
9.64
20.58
0.81
for votes that it disliked, it would be rather heroic to assume that that constitutes evidence
for the unidirectional influence of votes on aid. Finer-grained data on aid commitment dates
might help to answer that question; however, “Greenbook” only reports aid commitment on
an annual basis, so it is impossible to say which precedes which, the votes or the aid. In
order to obtain consistent estimates in light of this endogeneity, I set up a system of two
regression equations:
y1it = α0 + α1 y2it + αX 1it + γZ it + 1it
(1)
y2it = β0 + β1 y1it + βX 2it + δZ it + 2it
(2)
where y1 is a vector of logged aid from the US to country i in year t, y2 is a vector of the
percentage of times country i voted with the US in year t, X is the matrix of political and
economic controls discussed above, and Z is a vector of geographic controls.
Examining these equations, there are two reasons why endogeneity poses a problem.
First, the problem of reverse causality means that instrumental-variable estimation is needed
19
to obtain consistent estimates. Second, the disturbance terms of these two equations will
be correlated, which, if unattended to, will lead to inefficient estimates. A three-stage least
squares (3SLS) estimator allows me to take both of these issues into account and obtain
consistent estimates (Greene, 2003).
In order to properly identify a 3SLS model with two equations, it is necessary that each
equation in the system has at least one exclusion variable on the right hand side that does
not appear in the other model—that is, whose coefficient, were it in the other model, would
be zero. In the aid equation, that variable is the amount of aid from the US to other countries
in a given year. The assumption underlying this is that the US faces a budget constraint,
so the more aid it gives to one country, the less aid it has available for others. However, aid
given to others is unlikely to significantly impact how much a state votes with the US on the
Security Council. The one concern there may be in this regard is if we believe that states
care about relative amounts of aid received, rather than absolute. Thus, for example, if India
and Pakistan serve on the Security Council at the same time, and India sees Pakistan receive
greater aid, it may choose to retaliate against the US by voting against it more frequently
than it otherwise would.
This situation is unlikely to emerge, however, for two main reasons. First, in the example,
the Security Council seems like an unlikely venue for India’s protest. If the US is buying
Pakistan’s vote on the Security Council, chances are it is also buying India’s, so a defection
by India would likely lead to decreased aid to it, which is presumably something it would
like to avoid. Second, this type of situation is likely to only emerge when two countries
currently engaged in a dispute serve on the Council at the same time. Given that service on
the Council is infrequent, only lasts two years, and is subject to geographical restrictions, it
is unlikely that this would pose a major problem. And indeed, upon careful examination of
20
Council members in the observation period, I find that the only instances where this type of
situation might occur are during the three years that India and Pakistan have served on the
Council simultaneously. As a robustness check, I therefore test my models after omitting
these six observations, and find that excluding them does not change the results in any
meaningful way.4
The exclusion variable I use in the votes equations is a count variable of the number of
resolutions concerning Israel. It has been noted that the United States provides significant
diplomatic support for Israel within the United Nations, and often acts alone in protecting
Israel from reproach by the Security Council, mostly through the use of its veto power
(Mearsheimer and Walt, 2006). Due to the fact that the US frequently has to resort to its
veto power when it comes to draft resolutions on Israel, it seems clear that its views on this
aspect of Middle Eastern policy does not comport with that of other Council members, at
least not most of the time. Therefore, the greater the number of Israel resolutions voted on
by the Security Council, the more likely it is that the US is acting against the will of the
majority of the Council, which should lead to a decreased percentage of times that other
states vote with the US in a given year.
Furthermore, the selection of this variable as an instrument rests on the assumption that
the number of resolutions on Israel in a given year does not affect the amount of aid a country
receives from the US. While one might think that these resolutions are exactly the ones that
the US should be buying from other states, it is, in fact, rather unlikely that this would be
the case. The US knows full-well that its policies towards Israel are not shared by the vast
majority of states, and that it is unlikely to be able to reach consensus in the Council no
4
As an additional robustness check, I also try using US GDP as the exclusion variable in the aid equation
and get substantively the same results, with only very slight changes in the sizes of the coefficients of interest.
21
matter what it does. Thus, trying to buy the votes of Council members on Israel-related
resolutions is likely to be futile. Additionally, if the US were to try to buy favorable votes on
Israel, it would probably have to compromise on its position, at least to some extent, while
keeping the issue largely away from the Council’s purview means that it does not have to
deal with the negotiations that often go into the crafting of Council resolutions. Due to this,
it seems unlikely that the US would go to the effort of buying these particular votes on the
Security Council.
4
Results
The results of the simultaneous estimation, shown in Table 2, support the hypothesis that
increased levels of aid from the US lead to an increase in the amount of times that a state
votes with the US while on the Security Council. It does not seem to be the case, however,
that how states vote on the Council affects their aid receipts from the US. Although the
effect of votes on aid is not statistically significant at the traditional levels, both coefficients
are positive. The results from Model 2, my main regression model, furthermore support
the vote-buying mechanisms over the other two. The estimated effect of votes on aid shows
that an increase of one percent in the amount of times you vote with the US is associated
with an approximately 3 percent increase in the amount of aid received from the US. At
the same time, the effect of a 1 percent increase in aid from the US is associated with an
approximately 0.04 percent increase in the percentage of times a state votes with the US on
the Security Council in a given year.
The coefficients on the control variables in my models have the expected sign. Polity and
GDP are both negatively associated with the amount of aid received, which is as expected.
22
Table 2: Regression results
Variable
OLS
3SLS
OLS
Equation 1 : (log)Aid from US
Votes with US
0.026
0.030
0.030
-6e-04
(0.020)
(0.088)
(0.083)
∗∗∗
Polity
(log)GDP per capita
S score
(log)Population
US aid to others
Intercept
∗∗∗
-0.125
-0.126
-0.106
-0.101∗∗
(0.045)
(0.047)
(0.047)
(0.490)
-1.227
-1.228
0.494
0.538
(1.027)
(1.027)
(0.945)
(0.959)
13.190∗∗∗
13.164∗∗∗
(3.559)
(3.624)
-3.399∗∗∗
-3.367∗∗
-2.741∗∗
-3.045∗∗
(1.091)
(1.371)
(1.111)
(1.403)
-1e-04∗∗
-1e-04∗
-1e-04∗∗∗
-1e-04∗∗
(5e-05)
(6e-05)
(5e-05)
(6e-05)
48.736∗∗∗
48.225∗∗∗
16.438
36.618∗∗∗
41.338∗∗
16.482
(10.102)
R-squared
0.70
0.70
Equation 2: Percentage of votes with US
(log)Aid from US
0.442∗
3.596∗∗
(0.245)
Polity
(log)GDP per capita
S score
(log)Population
No. of Israel res.
Intercept
R-squared
N
Fixed Effects
Significance levels:
(0.021)
(1.690)
∗∗
3SLS
(9.860)
0.68
0.67
0.486∗∗
3.382∗∗
(0.235)
(1.514)
0.113
0.483
0.133
0.397
(0.166)
(0.298)
(0.163)
(0.255)
2.519
5.192
3.616
0.671
(3.592)
(5.077)
(3.191)
(4.499)
8.686
-36.096
(13.002)
(29.418)
-1.201
5.352
-0.701
3.228
(3.399)
(5.759)
(3.311)
(4.836)
∗∗∗
-0.882
-0.803
-0.864
-0.876∗∗
(0.262)
(0.358)
(0.261)
(0.346)
55.807
-50.537
46.737
-12.171
(26.679)
(66.591)
(22.936)
(42.861)
∗∗
∗∗
0.54
289
Country
∗ : 10%
∗∗ : 5%
23
0.14
289
Country
∗ ∗ ∗ : 1%
∗∗∗
0.53
289
Country
0.18
289
Yr/Ctry
At the same time, both variables are positively associated with how much a country votes
with the US, indicating that countries that are closer to the US on the polity scale (the US
is coded as 10 on the scale throughout the period of observation), and those that are more
developed tend to vote more with the US, regardless of other factors. The negative effect of
population in equation 1 is somewhat surprising, so, on the assumption that the observed
effect might be driven by a nonlinear relationship between population and aid, I add the
squared term of population into an unreported model as a robustness check. I find that,
indeed, the squared term is negative and significant while the linear term is positive and
insignificant, indicating a quadratic effect of population on aid. The size of my coefficients
of interest changes only slightly and the levels of statistical significance remain the same.
Interestingly, the S score seems to have the opposite impact from the endogenous variables. In equation 1, the S score has a statistically significant impact on the amount of aid
a country receives from the US, while votes are insignificant. In equation 2, however, the S
score does not have a statistically significant impact on how often states vote with the US
on the Security Council (p-value is 0.220), while votes do have such an impact. Although
the coefficient in equation 1 predicts a rather large average effect of the S score on aid (a
1316 percent increase in aid for a unit level increase in the S score), most changes in the
S score between years are at the second or third decimal point and a unit increase is never
observed in my dataset. Indeed, the mean change in the S score between two years for a
given country is a decrease of 0.003, which is associated with an average decrease in aid by
about 4 percent.
My two exclusion variables (US aid to others and No. of Israel Res.) in both equations
are in the posited direction and both are statistically significant.5 All my models include
5
The F statistic in the first stage regressions of the endogenous variables is highly significant (p-value
24
county fixed effects as additional controls, on the assumption that there is likely something
specific about each state on the Council that makes it more or less likely to vote with the
US.
6
Substantively, what these coefficients indicate is that a country that receives the mean
amount of aid from the US in a given year (USD 235 million) and votes with the US the
mean amount of times (87.19 percent) would need an approximate 30 percent increase in aid
in order to increase its voting with the US by about 1 percent.7 Depending on the number
of votes in that year, this change could buy the US either an abstention from the country on
a vote that the US cares about, or even an entire vote. If there is only one vote that the US
really cares about that year, a pivotal abstention may well be worth a 30 percent increase
in aid for one year (or even more).
For the purposes of comparison, Model 1 in Table 2 shows the results of an OLS regression
on the two dependent variables, i.e., an equation-by-equation estimation that does not take
the endogeneity of votes and aid into account. Note that while the coefficient of votes in
equation 1 remains largely the same (although the standard error in the simultaneous model
is over four times that of the OLS model), the effect of aid on votes (in equation 2), shows an
eight-fold increase and is statistically significant at the 5 percent level. As was noted above,
these results support the hypothesis that vote-buying may be taking place on the Security
is 0.00). The Anderson-Rubin, Likelihood-Ratio, Lagrange Multiplier, and conditional Wald test statistics
were calculated for both endogenous variables and all fail to reject the null hypothesis that the coefficient of
votes on aid is statistically different from zero at the 5% level, and all reject the null that the coefficient of
aid on votes is statistically different from zero at the 5% level.
6
Due to the lack of variation of the exclusion variable in the votes equation—the number of Israel-specific
resolutions per year in the Council—I am unable to control for both time- and country-specific effects in the
same regression. I also run my models with year-fixed effects rather than country-fixed effects. I find that
the magnitude of the coefficients on the two endogenous variables changes dramatically (the coefficient of
votes on aid changes from 0.030 to 0.459, and the coefficient of aid on votes changes from 3.596 to 0.385),
and that both are statistically significant in this specification.
7
A shift in aid of this magnitude is not unrealistic and, in fact, occurs several times in my dataset.
25
Council. However, it is harder to conclude which vote-buying mechanism is in operation.
This is made especially difficult by the fact that the effect of votes on aid is insignificant.
Had that coefficient been significant, it would render more likely the hypothesis that the
US bribes states for their votes and punishes them when they fail to comply. The lack of
a significant effect, however, does not necessarily contradict this hypothesis. Indeed, it is
possible that successful bribery is what leads to the insignificant impact of votes on aid—
namely, the vote-buying is successful, so the US rarely needs to respond to voting behavior
by punishing or rewarding other states for their behavior.
5
Conclusion
The empirical results above indicate that how states behave while serving as temporary members of the UN Security Council is not wholly divorced from the amount of aid they receive
from the United States. Thus, even though the relationship between Council membership
and aid that Kuziemko and Werker (2006) posited seems weaker than has been thought
until now, it is likely that the two are not entirely unrelated. In particular, my results show
that states respond to increases in aid from the United States, even though the US does not
respond to state voting behavior.
These results support the claim that the US is engaged in some type of vote-buying and
that the increased aid receipts observed during membership tenure on the Security Council do
not stem purely from other reasons. Indeed, if the “limelight” or “international involvement”
hypotheses were to hold, there should be no discernible effect of either endogenous variable on
the other. The question that remains, however, is which of the two vote-buying mechanisms
described above is the more likely one.
26
Although the finding that aid affects votes while votes do not affect aid does not directly
contradict the conventional hypothesis about bribery on the Security Council, it does little
to support it, either. Much has been made of anecdotal evidence of the US punishing states
that failed to vote the way the US wanted in particular instances. Indeed, the literature
has consistently used this evidence in combination with the impact of Security Council
membership on aid receipts from the US as sufficient evidence for bribery. And yet, while
scholars have largely identified that some kind of vote-buying takes place on the Security
Council, they are unable to identify its particular mechanism, and have not taken into
account the possibility that the US is buying votes by buying goodwill for itself. This is an
important distinction, as it speaks directly to US motives and the ways in which it tries to
influence world affairs.
These two vote-buying mechanisms are not easily distinguishable empirically. A significant effect of both endogenous variables on the other would lend support to the bribery
mechanism over the goodwill one, as there would then be evidence that the US seeks to
punish (or reward) states for their voting behavior. However, the absence of an empirical
effect of votes on aid does not mean that there is no bribery going on. Indeed, the Yemen
anecdote from above is simply one of several indicating that, under the right circumstances,
the US does not shy away from punishing states for their voting behavior.
With regard to the central question asked in this paper—whether the US engages in votebuying on the UN Security Council—the answer clearly points to the affirmative. With the
aid of recently available data on the voting behavior of temporary members of the Security
Council, I am able to distinguish between vote-buying and non-vote-buying behavior better
than previous literature has been able to. I furthermore posit that the existing literature
has insufficiently described the possible mechanisms through which vote-buying may take
27
place on the Council, namely bribing states for votes or buying their votes by buying their
goodwill. As was noted above, consensus is usually desired on the Council. In light of that,
it might, in fact, be cheaper for the US to buy goodwill among the elected members of the
Council than to engage in more vote-specific bribing. Additionally, the US does not want
to be caught engaging in outright bribery. Therefore, essentially buying everyone on the
Council off is a way to get close to the same result without having to be as concerned that
illicit or illegitimate behavior is uncovered. Indeed, if such behavior were to be exposed, that
might have a significantly detrimental impact on public opinion toward the organization, and
may reduce the legitimacy of, and support for, its operations. Thus, although this current
analysis cannot fully answer the question of which of these two mechanisms is the more likely
one, my empirical results, coupled with the likely incentives faced by the US, appear to favor
the goodwill mechanism.
28
References
Alesina, Alberto and David Dollar. 2000. “Who gives aid to whom and why?” Journal of
Economic Growth 5(1):33–63. 10
Bennett, D. Scott and Allan Stam. 2000. “EUGene: A Conceptual Manual.” International
Interactions 26:179–204. 18
Carter, David B. and Randall W. Stone. 2011. “Democracy and Multilateralism: The Case
of Vote Buying in the UN General Assembly.” Working Paper. 12
Chapman, Terrence L. and Dan Reiter. 2004. “The UN Security Council and the Rally
’Round-the-Flag Effect.” Journal of Conflict Resolution 48(6):886–909. 4
Chapman, Terrence L. and Scott Wolford. 2010. “International Organizations, Strategy, and
Crisis Bargaining.” Journal of Politics 72(1):227–242. 11
Dreher, Axel and James R. Vreeland. 2011. “Buying Votes and International Organizations.” Research Discussion Papers 123. Center for European Governance and Economic
Development. 4, 15, 16
Dreher, Axel, Jan-Egbert Sturm and James R. Vreeland. 2009a. “Development aid and
international politics: does membership on the UN Security Council influence World Bank
decisions?” Journal of Development Economics 88(1):1–18. 2, 7, 9, 10
Dreher, Axel, Jan-Egbert Sturm and James R. Vreeland. 2009b. “Global horse trading:
IMF loans for votes in the United Nations Security Council.” European Economic Review
53(7):742–757. 2, 7, 9, 15
Dreher, Axel, Peter Nunnenkamp and Rainer Thiele. 2008. “Does US aid buy UN General
Assembly votes?” Public Choice 136:139–164. 11
Fang, Songying. 2008. “The Informational Role of International Institutions and Domestic
Politics.” American Journal of Political Science 52(2):304–321. 4, 11
Gibler, Douglas M. and Meredith Sarkees. 2004. “Measuring Alliances: The Correlates
of War Formal Interstate Alliance Data set, 1816-2000.” Journal of Peace Research
41(2):211–222.
Greene, William H. 2003. Econometric Analysis, 5th ed. Upper Saddle River, NJ: Prentice
Hall. 20
Heston, Alan, Robert Summers and Bettina Aten. 2011. “Penn World Tables Version 7.0.”
http://pwt.econ.upenn.edu. 18
29
Hurd, Ian. 2002. “Legitimacy, power, and the symbolic life of the UN Security Council.”
Global Governance 8(1):35–51. 6
Kuziemko, Ilyana and Eric Werker. 2006. “How much is a seat on the seat on the Security
Council worth? Foreign aid and bribery at the United Nations.” Journal of Political
Economy 114(5):905–930. 2, 6, 7, 8, 9, 16, 26
Lowe, Vaughan, Adam Roberts, Jennifer Welsh and Dominik Zaum. 2008. The United
Nations Security Council and War: The Evolution of Thought and Practice since 1945.
New York: Oxford University Press. 6
Luck, Edward C. 2006. UN Security Council: Practice and Promise. New York, NY: Routledge. 6
Malone, David M. 2000. “Eyes on the Prize: the quest for nonpermanent seats on the UN
Security Council.” Global Governance 6(1):3–23. 6
Malone, David M. 2006a. The International Struggle Over Iraq: Politics in the UN Security
Council, 1980-2005. New York: Oxford University Press. 6
Malone, David M. 2006b. The UN Security Council: From the Cold War to the 21st Century.
New York, NY: Lynne Rienner Publishers. 6
McKinlay, Robert and Richard Little. 1977. “A foreign policy model of U.S. bilateral aid
allocations.” World Politics 30(1):58–86. 10
Mearsheimer, John J. and Stephen M. Walt. 2006. “The Israeli Lobby and U.S. Foreign
Policy.” Middle East Policy 13(3):29–87. 21
Monty G. Marshall, Ted Robert Gurr and Keith Jaggers. 2010. “POLITY IV PROJECT:
Political Regime Characteristics and Transitions, 1800-2009, Dataset Users Manual.”
http://www.systemicpeace.org/inscr/p4manualv2009.pdf. 18
Moreira, Marcelo J. and Brian P. Poi. 2003. “Implementing tests with correct size in simultaneous equations model.” Stata Journal 3(1):57–70.
Morgenthau, Hans. 1962. “A political theory of foreign aid.” American Political Science
Review 56(2):301–309. 10
Sandler, T and K Harley. 1999. The Political Economy of NATO: Past, Present, and into
the 21st Century. Cambridge: Cambridge University Press. 10
Stone, Randall W. 2011. Controlling Institutions: International Organizations and the Global
Economy. Cambridge: Cambridge University Press. 13
30
Thacker, Strom C. 1999. “The High Politics of IMF Lending.” World Politics 52:38–75. 16
United Nations. 1990. Security Council Resolution 678. Adopted November 29, 1990.
http://www.un.org/Docs/scres/1990/scres90.htm (July 28, 2012). 2
31